Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A vehicle with an internal combustion engine and a crankshaft for translating motion of pistons of the internal combustion engine into rotation, comprising: a firing level modulation module arranged to operate the internal combustion engine at a reduced effective displacement that is less than full displacement of the internal combustion engine by selectively modulating torque outputs of a cylinder, the modulated torque outputs including either or both of: (a) no torque output by intentionally skipping and not firing the cylinder; and (b) one or more torque output level(s) generated by controlling firing of the cylinder; and a machine learning module arranged to detect a misfire of the cylinder by learning to differentiate between an actual misfire of the cylinder versus either or both of: (c) the no torque output resulting from the intentional skipping and not firing of the cylinder; and (d) the one or more torque output level(s) resulting from controlling the firing of the cylinder.
This invention relates to internal combustion engines in vehicles, specifically addressing the challenge of accurately detecting misfires in engines that operate with variable displacement or torque modulation. Traditional misfire detection systems struggle to distinguish between intentional cylinder deactivation (for fuel efficiency) and actual misfires, leading to false positives or undetected faults. The invention solves this by integrating a firing level modulation module and a machine learning module. The firing level modulation module selectively reduces the engine's effective displacement by either skipping cylinder firings entirely or adjusting torque output levels, improving fuel efficiency. The machine learning module is trained to differentiate between intentional torque modulation (including skipped firings) and genuine misfires, ensuring accurate fault detection. By learning from engine operation patterns, the system avoids misclassifying controlled torque variations as misfires, enhancing diagnostic reliability. This approach enables precise misfire detection in engines with dynamic displacement control, improving both performance and emissions compliance.
2. The vehicle of claim 1 , wherein the machine learning module includes a neural network arranged to generate a predicted angular crank acceleration in response to receipt of one or more inputs indicative of operation of the vehicle in the firing level modulation mode.
This invention relates to vehicle control systems, specifically for managing engine operation in firing level modulation modes. The system addresses the challenge of accurately predicting engine behavior during transitions between different firing levels, which is critical for optimizing fuel efficiency and reducing emissions. The vehicle includes a machine learning module with a neural network that processes input data related to vehicle operation in firing level modulation mode. The neural network generates a predicted angular crank acceleration based on these inputs, enabling precise control of engine performance. The system dynamically adjusts engine parameters in response to the predicted acceleration, ensuring smooth transitions and minimizing inefficiencies. The neural network is trained to handle various operational conditions, improving adaptability and performance across different driving scenarios. This approach enhances engine responsiveness and reduces the risk of misfires or unstable combustion, contributing to overall vehicle efficiency and reliability. The invention focuses on leveraging machine learning to improve real-time engine control, particularly in modes where firing levels are dynamically adjusted.
3. The vehicle of claim 2 , wherein the one or more inputs are selected from: (a) a Fire Enable Flag for a next cylinder, a previous cylinder, and an opposing cylinder, respectively; (b) a cylinder skip number which is a number of skips preceding each firing for each of the plurality of cylinders in its own firing history respectively; (c) order skip number which is a number of skips preceding each firing in the firing order; (d) a Deceleration Cylinder Cut-Off (DCCO) exit flag which indicates an air pump-down event where valves associated with each of the cylinders are selectively opened without fuel injection to reduce intake manifold absolute pressure following DCCO events; (e) a weighted value compensating for effects of different skip fire patterns on angular crank acceleration; and (f) a fire skip status.
This invention relates to an engine control system for managing skip fire operation in internal combustion engines, particularly for optimizing fuel efficiency and performance. The system addresses the challenge of dynamically adjusting cylinder firing patterns to improve efficiency while maintaining smooth operation during deceleration and other transient conditions. The system processes multiple inputs to determine optimal firing sequences for each cylinder. These inputs include fire enable flags for adjacent and opposing cylinders, which control whether a cylinder should fire or skip. The system also tracks a cylinder skip number, representing the number of skips preceding each firing for individual cylinders, and an order skip number, indicating skips in the firing sequence. A deceleration cylinder cut-off (DCCO) exit flag signals air pump-down events, where valves are opened without fuel injection to reduce intake manifold pressure after deceleration. Additionally, a weighted value compensates for variations in angular crank acceleration caused by different skip fire patterns, and a fire skip status indicates whether a cylinder is currently firing or skipping. By integrating these inputs, the system dynamically adjusts firing sequences to enhance fuel efficiency, reduce emissions, and maintain drivability during transient conditions. The invention is particularly useful in engines employing skip fire operation to balance performance and efficiency.
4. The vehicle of claim 1 , wherein the machine learning module includes: a crank shaft acceleration module arranged to ascertain a measured crankshaft acceleration of the vehicle; a neural network which predicts a predicted angular acceleration of the crankshaft in response to a plurality of inputs indicative of operation of the vehicle in the firing level modulation mode; and a misfire detection module that determines when the misfire occurs by comparing the measured angular acceleration with the predicted angular acceleration of the crankshaft.
This invention relates to vehicle engine monitoring systems, specifically for detecting misfires in engines operating in a firing level modulation mode. The system addresses the challenge of accurately identifying misfires in engines where cylinder deactivation or partial combustion is used to improve fuel efficiency, as traditional misfire detection methods may not be reliable under these conditions. The system includes a machine learning module that processes real-time engine data to detect misfires. A crankshaft acceleration module measures the actual angular acceleration of the crankshaft. A neural network predicts the expected crankshaft acceleration based on multiple inputs related to engine operation, such as fuel injection timing, air intake, and ignition parameters. A misfire detection module compares the measured acceleration with the predicted acceleration. If the discrepancy exceeds a threshold, a misfire is detected. The neural network is trained to account for variations in engine behavior during firing level modulation, where some cylinders may be deactivated or operating at reduced power. This allows the system to distinguish between normal operational fluctuations and actual misfires. The approach improves detection accuracy in engines with variable combustion strategies, reducing false positives and ensuring reliable engine performance monitoring.
5. The vehicle of claim 4 , wherein the misfire detection module determines if the misfire has occurred using a misfire detection metric.
A vehicle monitoring system detects engine misfires by analyzing combustion events. The system includes sensors that measure engine parameters such as crankshaft speed, cylinder pressure, or exhaust gas composition. A misfire detection module processes these signals to identify incomplete or absent combustion in one or more cylinders. The module calculates a misfire detection metric, which quantifies deviations from expected combustion behavior. This metric may be derived from crankshaft speed fluctuations, pressure sensor readings, or exhaust gas oxygen levels. If the metric exceeds a predefined threshold, the system concludes that a misfire has occurred. The system may also track misfire frequency, location, and severity to assess engine health. This approach improves diagnostic accuracy by reducing false positives and ensuring timely detection of combustion irregularities. The system can trigger alerts, adjust engine operation, or log data for maintenance purposes. By continuously monitoring combustion events, the system enhances engine performance, reduces emissions, and prevents potential damage from prolonged misfires.
6. The vehicle of claim 5 , wherein the misfire detection metric is defined by: (a) generating a difference value by subtracting the measured angular acceleration from the predicted angular acceleration of the crankshaft; and (b) dividing the difference value by a normalized crankshaft acceleration value.
This invention relates to misfire detection in internal combustion engines, specifically improving the accuracy of detecting engine misfires by analyzing crankshaft angular acceleration. The problem addressed is the difficulty in reliably detecting misfires, particularly under varying operating conditions, due to noise and other disturbances affecting crankshaft acceleration measurements. The invention involves a vehicle equipped with an engine control system that calculates a misfire detection metric by comparing measured and predicted angular acceleration of the crankshaft. First, a difference value is generated by subtracting the measured angular acceleration from the predicted angular acceleration. The predicted angular acceleration is derived from engine operating parameters such as speed, load, and fuel injection timing. The measured angular acceleration is obtained from a crankshaft position sensor. The difference value is then divided by a normalized crankshaft acceleration value, which accounts for variations in engine dynamics under different conditions. This normalized metric enhances the sensitivity and reliability of misfire detection by reducing the impact of external noise and operational variability. The system may also include additional features such as adaptive filtering to refine the predicted acceleration or dynamic adjustment of the normalized value based on real-time engine conditions. The misfire detection metric is used to trigger corrective actions, such as adjusting ignition timing or fuel delivery, to mitigate the effects of misfires and improve engine performance and emissions. This approach provides a more robust and accurate method for identifying misfires compared to traditional techniques that rely solely on raw acceleration measuremen
7. The vehicle of claim 5 , wherein the misfire detection metric is normalized.
A vehicle monitoring system detects engine misfires by analyzing combustion events. The system includes sensors to measure engine parameters such as crankshaft speed, cylinder pressure, or exhaust gas composition. A controller processes these signals to identify misfires, which occur when a cylinder fails to ignite properly, leading to incomplete combustion and reduced engine efficiency. The system calculates a misfire detection metric based on the sensor data, which quantifies the likelihood or severity of misfires. To improve accuracy and comparability, this metric is normalized, meaning it is adjusted to account for variations in operating conditions such as engine speed, load, or temperature. Normalization ensures consistent detection thresholds across different driving scenarios, reducing false positives and negatives. The system may also compare the normalized metric to predefined thresholds to determine if a misfire has occurred and trigger appropriate responses, such as adjusting ignition timing or alerting the driver. This approach enhances engine performance, emissions control, and diagnostic reliability.
8. The vehicle of claim 5 , wherein misfire detection metric is dimensionless.
A vehicle monitoring system detects engine misfires by analyzing combustion events. The system includes sensors to measure engine parameters such as crankshaft speed, cylinder pressure, or exhaust gas composition. A processing unit calculates a misfire detection metric from these measurements, which quantifies the likelihood of a misfire occurring in one or more cylinders. The metric is dimensionless, meaning it is normalized or scaled to a unitless value for consistent comparison across different operating conditions. This allows the system to reliably identify misfires regardless of variations in engine speed, load, or environmental factors. The metric may be derived from statistical analysis, signal processing, or machine learning models trained on historical misfire data. The system may trigger diagnostic alerts, adjust engine control parameters, or log misfire events for maintenance purposes. By providing a dimensionless metric, the system ensures compatibility with various engine types and operating conditions, improving diagnostic accuracy and reducing false positives. The invention enhances engine performance, emissions control, and vehicle reliability by enabling precise misfire detection.
9. The vehicle of claim 5 , wherein misfire detection metric is amplified to increase a signal-to-noise ratio.
A vehicle monitoring system detects engine misfires by analyzing combustion events. The system measures variations in crankshaft rotational speed or other engine parameters to identify irregularities indicative of misfires. To improve detection accuracy, the system amplifies the misfire detection metric, which enhances the signal-to-noise ratio. This amplification process boosts the strength of the misfire-related signals relative to background noise, making it easier to distinguish genuine misfires from normal engine fluctuations. The system may use digital signal processing techniques, such as filtering or amplification algorithms, to refine the detection metric. By increasing the signal-to-noise ratio, the system reduces false positives and improves the reliability of misfire detection, ensuring timely maintenance and preventing engine damage. The amplified metric can be integrated into the vehicle's diagnostic system to trigger alerts or adjust engine control parameters as needed. This approach enhances the overall performance and longevity of the engine by providing more accurate and responsive misfire detection.
10. The vehicle of claim 5 , wherein the misfire detection module determines if the misfire has occurred based on the misfire detection metric exceeding a threshold.
A vehicle engine control system includes a misfire detection module that monitors engine operation to identify misfires, which are combustion events where fuel does not ignite properly. Misfires can lead to reduced engine performance, increased emissions, and potential damage to the catalytic converter. The system calculates a misfire detection metric, such as engine speed fluctuations or crankshaft acceleration, to assess combustion quality. The misfire detection module compares this metric against a predefined threshold to determine if a misfire has occurred. If the metric exceeds the threshold, the system flags the event as a misfire. The system may also include additional sensors, such as crankshaft position sensors, to provide input data for misfire detection. The threshold may be dynamically adjusted based on engine operating conditions, such as load, speed, or temperature, to improve detection accuracy. The system may further include diagnostic capabilities to log misfire events and trigger warning indicators for the driver. This approach ensures timely detection and response to misfires, enhancing engine reliability and emissions control.
11. The vehicle of claim 4 , wherein the neural network includes a plurality of hidden layers, each of the hidden layers includes one or more processors.
A vehicle includes a neural network with multiple hidden layers, each containing one or more processors. The neural network is trained to process sensor data from the vehicle's environment, such as images, LiDAR, or radar inputs, to perform tasks like object detection, path planning, or autonomous driving decisions. The hidden layers enable the neural network to extract hierarchical features from the sensor data, improving accuracy in identifying and classifying objects, predicting movements, or making real-time driving decisions. The processors in each hidden layer execute computations to transform input data through successive layers, refining the output for higher-level decision-making. This architecture allows the neural network to handle complex environmental data efficiently, supporting autonomous vehicle operations by interpreting surroundings, avoiding obstacles, and navigating routes safely. The system may integrate with other vehicle components, such as control systems or user interfaces, to execute actions based on the neural network's outputs. The use of multiple hidden layers enhances the network's ability to learn from large datasets, improving performance in dynamic driving conditions.
12. The vehicle of claim 11 , wherein the one or more processors use a hyperbolic tangent activation function.
A system for vehicle control utilizes one or more processors to implement a neural network for processing sensor data and generating control signals. The neural network includes at least one layer with a hyperbolic tangent activation function, which helps the network model complex, non-linear relationships in the data. The system processes input data from various sensors, such as cameras, LiDAR, or radar, to detect objects, predict trajectories, and determine optimal control actions. The hyperbolic tangent function, defined as tanh(x) = (e^x - e^-x)/(e^x + e^-x), provides a smooth, differentiable output ranging between -1 and 1, which improves training stability and convergence. This activation function is particularly useful in layers where the network must handle a wide range of input values while maintaining precise control over output dynamics. The system may be integrated into autonomous vehicles, advanced driver-assistance systems (ADAS), or other vehicle control applications where real-time decision-making is critical. The use of hyperbolic tangent activation enhances the network's ability to learn and adapt to varying driving conditions, improving overall system performance and safety.
13. The vehicle of claim 11 , wherein the one or more processors are arranged to learn based on training data set using a Limited-memory BFGS (L-BFGS) technique.
This invention relates to a vehicle equipped with a machine learning system for optimizing performance. The vehicle includes one or more processors configured to process sensor data and execute a machine learning algorithm. The system is trained using a Limited-memory BFGS (L-BFGS) optimization technique, which is a quasi-Newton method for large-scale optimization problems. L-BFGS is particularly suited for training machine learning models by approximating the Hessian matrix, reducing computational complexity while maintaining accuracy. The vehicle's processors use this trained model to analyze sensor inputs, such as environmental conditions, vehicle dynamics, or user behavior, to make real-time decisions. These decisions may include adjusting vehicle parameters, predicting maintenance needs, or enhancing autonomous driving capabilities. The L-BFGS technique allows the system to efficiently handle large datasets and complex models, improving the vehicle's adaptability and performance. The invention addresses the challenge of optimizing machine learning models in resource-constrained environments, such as vehicles, by leveraging efficient optimization algorithms.
14. The vehicle of claim 1 , wherein the machine learning module includes a neural network arranged to generate a misfire flag in response to receipt of (a) a measured angular crank acceleration and (b) one or more inputs indicative of operation of the vehicle in the firing level modulation mode.
This invention relates to vehicle engine control systems, specifically addressing the detection of engine misfires during operation in a firing level modulation mode. The firing level modulation mode is a strategy used to reduce engine output by selectively deactivating some cylinders, improving fuel efficiency under certain conditions. However, this mode can complicate traditional misfire detection methods, as the expected engine behavior differs from normal operation. The system includes a machine learning module with a neural network trained to identify misfires by analyzing a measured angular crank acceleration signal. The neural network processes this signal alongside additional inputs that indicate the vehicle is operating in the firing level modulation mode. These inputs may include cylinder deactivation patterns, throttle position, or other operational parameters that influence engine behavior. By incorporating these inputs, the neural network can distinguish between normal firing level modulation behavior and actual misfires, improving detection accuracy. The system generates a misfire flag when a misfire is detected, which can trigger diagnostic alerts or corrective actions. This approach enhances reliability in misfire detection during partial-cylinder operation, ensuring compliance with emissions regulations and preventing engine damage.
15. The vehicle of claim 14 , wherein the machine learning module is further arranged to generate the misfire flag based on a probability score generated by the neural network, the probability score derived from a comparison between the measured angular crank acceleration and a predicted angular acceleration of the crankshaft.
This invention relates to vehicle engine monitoring systems that detect misfires using machine learning. The problem addressed is the need for accurate and reliable misfire detection to prevent engine damage and reduce emissions. Traditional methods often rely on fixed thresholds or simple algorithms, which can be less effective under varying operating conditions. The system includes a machine learning module that processes sensor data, particularly angular crank acceleration measurements from the engine's crankshaft. The module uses a neural network to analyze these measurements and compare them against predicted angular acceleration values. The neural network generates a probability score indicating the likelihood of a misfire event. If this score exceeds a predefined threshold, a misfire flag is triggered, alerting the vehicle's control system to take corrective action. The neural network is trained on historical data to improve accuracy, adapting to different engine conditions and operating states. This approach enhances detection sensitivity and reduces false positives compared to traditional methods. The system can be integrated into existing engine control units or as a standalone diagnostic tool, providing real-time monitoring and feedback to optimize engine performance and longevity.
16. The vehicle of claim 15 , wherein the neural network generates the predicted angular crank acceleration of the crankshaft from the one or more inputs indicative of operation of the vehicle in the firing level modulation mode.
This invention relates to vehicle engine control systems, specifically for predicting crankshaft angular acceleration in firing level modulation (FLM) mode. FLM mode is used to reduce engine torque output by selectively deactivating some cylinders, which creates irregular combustion patterns that can affect crankshaft dynamics. The challenge is accurately predicting crankshaft behavior under these conditions to optimize engine performance and emissions. The system uses a neural network trained to process multiple inputs related to vehicle operation during FLM mode. These inputs may include engine speed, throttle position, cylinder activation patterns, and other operational parameters. The neural network analyzes these inputs to generate a predicted angular acceleration of the crankshaft, accounting for the irregular torque fluctuations caused by FLM. This prediction helps the engine control unit adjust fuel injection, ignition timing, or other parameters to maintain smooth operation and minimize vibrations. The neural network is specifically configured to handle the unique challenges of FLM mode, where traditional physics-based models may struggle due to the non-linear and intermittent combustion events. By using machine learning, the system can adapt to different driving conditions and engine configurations more effectively than conventional approaches. The predicted crankshaft acceleration can be used for various control strategies, including torque smoothing, noise reduction, and improved fuel efficiency.
17. The vehicle of claim 14 , wherein the neural network includes a plurality of hidden layers, each of the plurality of hidden layers having at least one processor.
A system for autonomous vehicle control uses a neural network with multiple hidden layers to process sensor data and generate control signals for vehicle navigation. Each hidden layer in the neural network is implemented with at least one dedicated processor, enabling parallel processing of data to improve computational efficiency and decision-making speed. The neural network is trained to interpret inputs from various sensors, such as cameras, LiDAR, and radar, to detect obstacles, road conditions, and other vehicles. The system then generates outputs that control steering, acceleration, and braking to navigate the vehicle safely. The use of multiple processors in each hidden layer allows for distributed computation, reducing latency and enhancing real-time performance. This architecture supports complex decision-making tasks, such as path planning and collision avoidance, by leveraging the parallel processing capabilities of the neural network. The system may also integrate additional layers or processors to handle specific tasks, such as object recognition or environmental mapping, further improving the vehicle's autonomy and safety. The neural network's design ensures scalability, allowing for the addition of more layers or processors as computational demands increase. This approach enhances the vehicle's ability to operate in dynamic environments while maintaining high reliability and performance.
18. The vehicle of claim 17 , wherein the at least one processor performs a Rectified Linear (“ReLU”) activation function.
A system for processing data in a vehicle includes at least one processor configured to execute a neural network model. The neural network model is trained to perform a specific task, such as object detection, classification, or decision-making, using input data from vehicle sensors. The system enhances the neural network's performance by applying a Rectified Linear Unit (ReLU) activation function during processing. ReLU is a mathematical function that outputs the input directly if it is positive; otherwise, it outputs zero. This function helps mitigate the vanishing gradient problem in deep neural networks, improving training efficiency and model accuracy. The processor may also perform other neural network operations, such as convolution, pooling, or fully connected layers, to extract and analyze features from the input data. The system is designed to operate in real-time, enabling the vehicle to make autonomous decisions based on processed sensor inputs. The use of ReLU activation ensures faster convergence during training and enhances the model's ability to generalize from training data to real-world scenarios. The overall system improves the reliability and efficiency of autonomous vehicle operations by leveraging advanced neural network techniques.
19. The vehicle of claim 17 , wherein the at least one processor is arranged to learn from training data using a Stochastic Gradient Descent (SGD) technique.
This invention relates to a vehicle equipped with a machine learning system for processing sensor data. The system includes at least one processor configured to receive sensor data from one or more sensors mounted on the vehicle, such as cameras, LiDAR, or radar. The processor analyzes this data to detect and classify objects in the vehicle's environment, such as pedestrians, other vehicles, or road signs. The system further includes a memory storing a machine learning model trained to perform these tasks. The processor executes the model to generate outputs, such as object classifications or control signals for autonomous driving functions. A key feature is the processor's ability to learn from training data using a Stochastic Gradient Descent (SGD) technique. SGD is an iterative optimization algorithm that adjusts the model's parameters to minimize prediction errors. The processor updates the model incrementally, processing small batches of training data at a time. This approach improves the model's accuracy over time by refining its ability to interpret sensor inputs and make decisions, such as adjusting vehicle speed or steering. The system may also include additional processors or modules for real-time data processing, ensuring timely responses to dynamic driving conditions. The overall goal is to enhance the vehicle's perception and decision-making capabilities for safer and more efficient autonomous operation.
20. The vehicle of claim 1 , wherein the firing level modulation mode is a skip fire mode, wherein a select cylinder is fired, skipped and selectively either fired or skipped in successive working cycles while the internal combustion engine is operating at the reduced effective displacement.
This invention relates to internal combustion engines, specifically systems and methods for modulating firing levels to improve efficiency and performance. The technology addresses the problem of excessive fuel consumption and emissions during partial load operation by selectively deactivating certain cylinders in a skip fire mode. In this mode, a specific cylinder is either fired or skipped in successive working cycles, allowing the engine to operate at a reduced effective displacement while maintaining smooth operation. The system dynamically adjusts the firing sequence based on engine demand, optimizing fuel efficiency and reducing unnecessary combustion events. This approach differs from traditional cylinder deactivation, which fully deactivates cylinders for extended periods, by allowing more granular control over individual cylinder firing. The skip fire mode ensures that only the necessary cylinders are fired to meet power requirements, minimizing energy waste and emissions. The invention is particularly useful in vehicles where fuel economy and emissions reduction are critical, such as hybrid or high-efficiency internal combustion engines. The system may include sensors and control logic to monitor engine conditions and adjust the firing sequence in real-time, ensuring optimal performance under varying operating conditions.
21. The vehicle of claim 20 , wherein the machine learning module includes a neural network arranged to receive at least one skip fire operating parameter that is characteristic of the skip fire mode of operation of the internal combustion engine, the at least one skip fire operating parameter including one or more of the following: (a) a Fire Enable Flag for a next cylinder, a previous cylinder, and an opposing cylinder, respectively; (b) a cylinder skip number which is a number of skips preceding each firing for each of a plurality of cylinders in its own firing history respectively; (c) order skip number which is a number of skips preceding each firing in the firing order; (d) a Deceleration Cylinder Cut-Off (DCCO) exit flag which indicates an air pump-down event where valves associated with each of the cylinders are selectively opened without fuel injection to reduce intake manifold absolute pressure following DCCO events; (e) a weighted value compensating for effects of different skip fire patterns on angular crank acceleration; and (f) a skip fire status.
This invention relates to internal combustion engines operating in skip fire mode, where cylinders are selectively activated or deactivated to optimize fuel efficiency and performance. The challenge addressed is accurately modeling and controlling skip fire operation to improve engine responsiveness and reduce emissions. The system includes a machine learning module with a neural network that processes skip fire operating parameters to enhance control decisions. These parameters include a Fire Enable Flag for adjacent and opposing cylinders, a cylinder skip number tracking skips before each firing for individual cylinders, an order skip number tracking skips in the firing sequence, a Deceleration Cylinder Cut-Off (DCCO) exit flag indicating air pump-down events where valves are opened without fuel injection to reduce intake manifold pressure, a weighted value compensating for varying skip fire patterns' effects on crank acceleration, and a skip fire status indicator. The neural network uses these inputs to refine skip fire control, improving engine efficiency and reducing torque disturbances during transitions. The system dynamically adapts to different driving conditions by analyzing historical firing patterns and adjusting control strategies accordingly. This approach enhances skip fire operation by leveraging machine learning to optimize cylinder activation sequences and mitigate performance variations caused by skip fire patterns.
22. The vehicle of claim 1 , wherein the machine learning module includes a neural network which is arranged to receive one or more inputs indicative of operation of the vehicle, the one or more inputs including: spark timing; fuel mass; fire skip status; fire enable flag; cylinder skip number; order skip number; mass air per cylinder; cam phase timing; charge air temperature; engine speed; manifold absolute pressure; transmission gear; Deceleration Cylinder Cut-Off (DCCO) exit; vehicle speed; torque request; pedal position; fuel pressure; and turbocharger waste gate position.
This invention relates to a vehicle equipped with a machine learning module for optimizing engine performance. The system addresses the challenge of improving fuel efficiency, emissions control, and drivability by dynamically adjusting engine parameters based on real-time operational data. The machine learning module uses a neural network trained to process multiple inputs related to vehicle operation, including spark timing, fuel mass, fire skip status, fire enable flag, cylinder skip number, order skip number, mass air per cylinder, cam phase timing, charge air temperature, engine speed, manifold absolute pressure, transmission gear, Deceleration Cylinder Cut-Off (DCCO) exit, vehicle speed, torque request, pedal position, fuel pressure, and turbocharger waste gate position. The neural network analyzes these inputs to generate optimized control signals for engine components, enabling adaptive adjustments to enhance performance. The system may also incorporate additional features such as predictive maintenance, fault detection, and real-time calibration updates to further refine engine operation. By leveraging machine learning, the vehicle achieves more precise and efficient control over engine parameters, reducing fuel consumption and emissions while maintaining optimal power delivery.
23. The vehicle of claim 1 , further comprising a misfire counter arranged to count a plurality of misfires as determined by the machine learning module and to generate a notification when the plurality of misfires exceeds a threshold value.
This invention relates to vehicle engine monitoring systems that use machine learning to detect and analyze engine misfires. The system includes a machine learning module trained to identify misfire events based on engine sensor data, such as crankshaft speed fluctuations or cylinder pressure readings. The module processes real-time sensor inputs to classify normal and abnormal combustion events, distinguishing between different types of misfires (e.g., partial or complete). A misfire counter tracks the frequency of these events and compares them against a predefined threshold. When the misfire count exceeds the threshold, the system generates a notification to alert the driver or vehicle diagnostics system, indicating potential engine issues. The notification may trigger maintenance alerts or adjustments to engine control parameters to mitigate further damage. The system improves upon traditional misfire detection methods by using machine learning to enhance accuracy and adapt to varying engine conditions, reducing false positives and improving diagnostic reliability. This approach helps prevent long-term engine damage and ensures optimal performance by proactively addressing misfire-related problems.
24. The vehicle of claim 1 , wherein the one or more torque output levels are each generated by using different air charge and/or fueling levels when firing the at least one cylinder.
This invention relates to vehicle engine control systems designed to optimize torque output through variable air charge and fueling levels. The problem addressed is the need for precise torque modulation in internal combustion engines to improve performance, efficiency, and emissions control. The system adjusts torque output by selectively firing one or more cylinders with different air charge and fueling levels, allowing for fine-grained control over engine power delivery. By varying these parameters, the engine can achieve multiple torque output levels without requiring mechanical or structural changes. This approach enables smoother transitions between torque settings, reduces fuel consumption, and minimizes emissions. The invention is particularly useful in applications where rapid and precise torque adjustments are necessary, such as hybrid vehicles or engines with advanced combustion strategies. The system may also incorporate feedback mechanisms to dynamically adjust air charge and fueling based on real-time operating conditions, ensuring optimal performance across different driving scenarios. This method enhances engine responsiveness while maintaining efficiency and compliance with emissions regulations.
25. The vehicle of claim 1 , wherein the firing level modulation mode is a skip fire mode wherein the plurality of cylinders are fired and skipped in a predefined rolling pattern.
This invention relates to engine control systems, specifically for modulating the firing levels of internal combustion engines to improve efficiency and reduce emissions. The problem addressed is the need for more precise control over engine operation to optimize fuel consumption and performance under varying driving conditions. The system includes an engine with multiple cylinders and a controller that adjusts the firing sequence of these cylinders. In a skip fire mode, the controller selectively fires and skips cylinders in a predefined rolling pattern, where the pattern cycles through different combinations of active and inactive cylinders. This approach allows the engine to operate at varying torque levels while maintaining smooth operation and reducing unnecessary fuel consumption. The rolling pattern ensures that no single cylinder is consistently overused, promoting even wear and extending engine life. The controller may also adjust the firing sequence based on real-time operating conditions, such as load demand or speed, to further optimize efficiency. By dynamically modulating the firing levels, the system achieves better fuel economy and reduced emissions compared to traditional fixed firing strategies. The skip fire mode is particularly useful in hybrid or variable-displacement engines where partial cylinder deactivation is beneficial.
26. The vehicle of claim 1 , wherein the machine learning module is classification based and is capable of predicting misfire flags.
A system for vehicle engine monitoring uses a machine learning module to detect engine misfires. The module is classification-based, meaning it categorizes engine operation data into predefined classes, such as normal operation or misfire conditions. By analyzing sensor inputs like engine speed, exhaust gas composition, and ignition timing, the module predicts misfire events in real-time. This helps improve engine performance, reduce emissions, and prevent damage by identifying misfires early. The system integrates with the vehicle's onboard diagnostics to flag misfires for further action. The machine learning model is trained on historical engine data to recognize patterns associated with misfires, ensuring accurate predictions. This approach enhances traditional misfire detection methods by leveraging advanced data analysis techniques. The system is particularly useful in modern vehicles where precise engine control and emissions compliance are critical. By continuously monitoring and classifying engine behavior, the module enables proactive maintenance and optimizes engine efficiency. The technology addresses the challenge of detecting misfires in complex engine systems where traditional sensors may not provide sufficient accuracy.
27. An engine controller arranged to control operation of an internal combustion engine of a vehicle, the engine controller comprising: a firing level modulation module arranged to operate the internal combustion engine in a firing level modulation mode such that an output of a select cylinder, during a firing opportunity, is modulated so that the select cylinder selectively is either (a) intentionally skipped so as to generate no torque during the firing opportunity or (b) controlled so as to generate one or more different torque output levels when fired during the firing opportunity; and a machine learning module arranged to detect a misfire of the select cylinder during the firing opportunity by differentiating between an actual misfire of the select cylinder versus the intentionally skipping or generation of the one or more different torque output levels respectively.
This invention relates to engine control systems for internal combustion engines, specifically addressing the challenge of accurately detecting cylinder misfires in engines that use firing level modulation. Firing level modulation involves selectively skipping cylinder firings or adjusting torque output levels during firing opportunities to optimize engine performance. However, this technique can interfere with traditional misfire detection methods, as intentional skips or torque variations may be mistaken for actual misfires. The engine controller includes a firing level modulation module that operates the engine in a firing level modulation mode. In this mode, the output of a selected cylinder is modulated during a firing opportunity, either by intentionally skipping the firing (generating no torque) or by controlling the cylinder to produce one or more different torque levels when fired. The controller also includes a machine learning module designed to distinguish between actual misfires and the intentional skips or torque variations. This differentiation ensures that misfire detection remains accurate even when firing level modulation is active, preventing false positives or missed misfires. The machine learning module analyzes engine behavior to identify true misfires while accounting for the deliberate changes in cylinder operation. This approach improves engine diagnostics and reliability by maintaining accurate misfire detection in advanced engine control strategies.
28. The engine controller of claim 27 , wherein the machine learning module includes a misfire detection module that determines when the misfire occurs by comparing a measured angular acceleration with a predicted angular acceleration of the crankshaft.
This invention relates to engine control systems, specifically improving misfire detection in internal combustion engines. The problem addressed is the need for accurate and reliable misfire detection to prevent engine damage and reduce emissions. Traditional methods often rely on sensor data analysis but may lack precision or require extensive computational resources. The engine controller includes a machine learning module that enhances misfire detection. This module contains a misfire detection submodule that compares measured angular acceleration of the crankshaft with a predicted angular acceleration. The predicted acceleration is generated using a model trained on engine operating conditions, allowing real-time detection of deviations indicative of misfires. The comparison process involves analyzing discrepancies between the measured and predicted values, where significant deviations trigger a misfire alert. The system may also adjust engine parameters in response to detected misfires to mitigate performance degradation. The machine learning approach improves detection accuracy and reduces false positives compared to conventional methods. This solution is particularly useful in modern engines where precise control and diagnostics are critical for efficiency and emissions compliance.
29. The engine controller of claim 28 , wherein the misfire detection module is arranged to receive the measured crankshaft acceleration signal from a crank acceleration calculation module provided on the vehicle.
This invention relates to engine control systems, specifically focusing on misfire detection in internal combustion engines. The problem addressed is the need for accurate and reliable misfire detection to ensure engine performance, emissions compliance, and drivability. Misfires occur when one or more cylinders fail to ignite properly, leading to incomplete combustion and potential damage to the engine or catalytic converter. The invention describes an engine controller that includes a misfire detection module designed to analyze crankshaft acceleration signals to identify misfires. The misfire detection module receives a measured crankshaft acceleration signal from a crank acceleration calculation module installed on the vehicle. The crank acceleration calculation module processes raw sensor data, such as crankshaft position or speed, to derive the acceleration signal, which reflects the rotational dynamics of the engine. By analyzing fluctuations in this signal, the misfire detection module can determine whether a misfire has occurred in one or more cylinders. The system enhances diagnostic accuracy by leveraging real-time crankshaft acceleration data, allowing for timely detection and mitigation of misfires. This improves engine efficiency, reduces emissions, and prevents long-term damage. The invention is particularly useful in modern vehicles where precise engine control and emissions monitoring are critical.
30. The engine controller of claim 28 , further comprising a neural network which determines the predicted angular acceleration of the crankshaft in response to a plurality of inputs indicative of operation of the vehicle.
This invention relates to engine control systems, specifically an engine controller that improves crankshaft angular acceleration prediction for vehicle operation. The problem addressed is the need for accurate and responsive crankshaft acceleration estimation to optimize engine performance, fuel efficiency, and emissions control. The engine controller includes a neural network that processes multiple inputs related to vehicle operation to predict the crankshaft's angular acceleration. These inputs may include engine speed, throttle position, fuel injection parameters, ignition timing, and other relevant sensor data. The neural network is trained to analyze these inputs and output a predicted angular acceleration value, which can be used to adjust engine control strategies in real time. The neural network enhances the controller's ability to anticipate changes in crankshaft dynamics, allowing for more precise control of fuel delivery, ignition timing, and other engine functions. This improves overall engine efficiency, reduces emissions, and enhances drivability. The system may also incorporate additional features, such as adaptive learning to refine predictions over time based on actual engine performance data. By leveraging machine learning, the controller provides a more accurate and adaptive solution compared to traditional physics-based models, particularly in dynamic driving conditions. The invention is applicable to various internal combustion engines, including those in automotive, marine, and industrial applications.
31. The engine controller of claim 30 , wherein the one or more inputs are selected from: (a) a Fire Enable Flag for a next cylinder, a previous cylinder, and an opposing cylinder, respectively; (b) a cylinder skip number which is a number of skips preceding each firing for each of the plurality of cylinders in its own firing history respectively; (c) order skip number which is a number of skips preceding each firing for each of the plurality of cylinders in the firing order respectiyely.
This invention relates to engine control systems, specifically for managing cylinder firing sequences in internal combustion engines. The problem addressed is optimizing engine performance by selectively enabling or disabling cylinder firing based on dynamic operating conditions, such as fuel efficiency, emissions reduction, or power output requirements. The engine controller monitors and adjusts cylinder firing using multiple inputs to determine firing sequences. These inputs include Fire Enable Flags for individual cylinders, which indicate whether a cylinder is permitted to fire in its next, previous, or opposing cylinder's firing cycle. Additionally, the controller tracks a cylinder skip number, representing the number of skips preceding each firing event for each cylinder in its own firing history. This allows the system to analyze past firing patterns to optimize future firing decisions. The controller also considers an order skip number, which represents the number of skips preceding each firing event for each cylinder in the engine's firing order sequence. This helps synchronize firing decisions across multiple cylinders to maintain engine stability and performance. By integrating these inputs, the engine controller dynamically adjusts firing sequences to improve efficiency, reduce emissions, or enhance power output based on real-time operating conditions. The system ensures smooth engine operation by balancing individual cylinder performance with overall engine coordination.
32. The engine controller of claim 28 , wherein the misfire detection module determines if the misfire has occurred based on a misfire detection metric exceeding a threshold.
The invention relates to engine control systems, specifically focusing on misfire detection in internal combustion engines. Misfires occur when a cylinder fails to ignite properly, leading to incomplete combustion, reduced efficiency, and increased emissions. Accurate misfire detection is critical for engine performance, emissions control, and diagnostic systems. The engine controller includes a misfire detection module that evaluates engine operation to identify misfires. The module calculates a misfire detection metric, which quantifies deviations in engine behavior indicative of misfires. This metric is compared to a predefined threshold. If the metric exceeds the threshold, the system concludes that a misfire has occurred. The threshold may be dynamically adjusted based on operating conditions to improve detection accuracy. The misfire detection metric may be derived from various engine parameters, such as crankshaft speed fluctuations, ion sensing signals, or cylinder pressure measurements. The system may also incorporate additional logic to distinguish between true misfires and transient events, reducing false positives. Once a misfire is detected, the controller may trigger corrective actions, such as adjusting ignition timing, fuel delivery, or alerting the driver. This approach enhances engine reliability by ensuring timely misfire identification, which is essential for maintaining compliance with emissions regulations and optimizing fuel efficiency. The system is particularly useful in modern engines where precise control and diagnostics are required.
33. The engine controller of claim 32 , wherein the misfire detection metric is at least one of the following (a) normalized, (b) dimensionless and (c) amplified to increase a signal-to-noise ratio.
This invention relates to engine control systems, specifically improving misfire detection in internal combustion engines. The problem addressed is the difficulty in accurately detecting engine misfires due to noise and signal variability, which can lead to false positives or missed detections, affecting engine performance and emissions. The engine controller includes a misfire detection system that processes engine speed fluctuations to identify misfires. The system calculates a misfire detection metric, which is then enhanced to improve reliability. The metric can be normalized to account for variations in engine operating conditions, making it dimensionless for consistent comparison. Additionally, the metric can be amplified to boost the signal-to-noise ratio, ensuring that misfire events are more distinguishable from background noise. These enhancements allow the controller to more accurately detect misfires, even under challenging conditions, improving diagnostic accuracy and engine control. The system may also include adaptive thresholds or filtering to further refine misfire detection. By processing the metric in these ways, the controller reduces false alarms and ensures timely detection of misfires, supporting better engine management and emissions control. This approach is particularly useful in modern engines where precise diagnostics are critical for efficiency and regulatory compliance.
34. The engine controller of claim 27 , wherein the machine learning module includes a neural network arranged to detect the misfire in response to receipt of: (a) a measured angular crank acceleration input; and (b) a plurality of inputs indicative of operation of the vehicle.
This invention relates to engine control systems that use machine learning to detect engine misfires. The problem addressed is the need for accurate and reliable misfire detection in vehicles, which is critical for engine performance, emissions control, and diagnostic systems. Traditional misfire detection methods often rely on fixed thresholds or rule-based algorithms, which may not adapt well to varying operating conditions or different engine types. The invention involves an engine controller with a machine learning module that includes a neural network. The neural network is trained to detect misfires by analyzing two key inputs: (1) a measured angular crank acceleration signal, which reflects the rotational dynamics of the engine's crankshaft, and (2) multiple inputs indicative of vehicle operation, such as engine speed, load, throttle position, and other relevant parameters. By processing these inputs, the neural network can identify misfire events with improved accuracy compared to conventional methods. The machine learning approach allows the system to adapt to different driving conditions and engine characteristics, enhancing detection reliability. This method is particularly useful in modern engines where misfire detection must be precise to meet stringent emissions and performance standards. The neural network may be trained using historical data or real-time learning techniques to continuously improve detection performance.
35. The engine controller of claim 34 , wherein the neural network is further configured to: (c) generate a predicted angular crank acceleration from the plurality of inputs indicative of the operation of the vehicle; (d) generate a misfire flag based on a misfire detection metric exceeding a predefined threshold, the misfire detection metric derived from a comparison of the measured angular crank acceleration input and the predicted angular crank acceleration.
This invention relates to engine control systems that use neural networks to detect engine misfires. The problem addressed is the need for accurate and reliable misfire detection in internal combustion engines to ensure optimal performance, reduce emissions, and prevent damage. Traditional misfire detection methods often rely on fixed thresholds or simplified models, which may not adapt well to varying operating conditions or engine states. The engine controller includes a neural network trained to process multiple inputs indicative of vehicle operation, such as crankshaft position, engine speed, and other sensor data. The neural network generates a predicted angular crank acceleration based on these inputs. The system compares this predicted acceleration with the measured angular crank acceleration from the engine. If the difference between the predicted and measured values exceeds a predefined threshold, a misfire detection metric is calculated. When this metric surpasses a set threshold, a misfire flag is generated, indicating a misfire event. This approach improves detection accuracy by leveraging the neural network's ability to model complex relationships between engine parameters and misfire conditions. The system dynamically adapts to changing conditions, enhancing reliability compared to traditional methods.
36. The engine controller of claim 34 , wherein the neural network is further configured to: generate a misfire flag based on a probability score exceeding a predetermined value, wherein the probability score is an output of the neural network.
This invention relates to engine control systems that use neural networks to detect engine misfires. Engine misfires occur when fuel combustion fails to ignite properly, leading to reduced efficiency, increased emissions, and potential damage. Traditional misfire detection methods rely on engine speed fluctuations or exhaust gas analysis, which can be inaccurate or slow to respond. The invention improves upon these methods by using a neural network to analyze engine sensor data and predict misfires with higher accuracy. The neural network processes input data from engine sensors, such as crankshaft position, combustion pressure, and exhaust gas composition. It generates a probability score indicating the likelihood of a misfire. If this score exceeds a predetermined threshold, the neural network outputs a misfire flag, which the engine controller uses to adjust fuel injection, ignition timing, or other parameters to correct the issue. The neural network is trained on historical engine data to improve its accuracy over time. This approach provides faster and more reliable misfire detection compared to conventional methods, enhancing engine performance and reducing emissions. The system can be integrated into existing engine control units with minimal hardware modifications.
37. The engine controller of claim 27 , wherein the machine learning module includes a neural network having a plurality of hidden layers, each of the hidden layers having one or more processors.
This invention relates to an engine controller incorporating a machine learning module with a neural network for improved engine performance and efficiency. The neural network includes multiple hidden layers, each containing one or more processors to enhance computational capabilities. The engine controller monitors engine parameters such as temperature, pressure, and fuel consumption in real-time. The neural network processes these inputs through its hidden layers to generate optimized control signals for engine components like fuel injectors, turbochargers, and exhaust systems. This adaptive learning approach allows the controller to adjust engine operations dynamically, improving fuel efficiency, reducing emissions, and extending engine lifespan. The neural network's multi-layered architecture enables complex pattern recognition and decision-making, outperforming traditional rule-based control systems. The processors within each hidden layer handle parallel computations, accelerating response times and enabling real-time adjustments. This advanced control strategy addresses challenges in engine performance variability, environmental regulations, and operational efficiency in automotive and industrial applications. The invention focuses on integrating deep learning techniques into engine management systems to achieve superior adaptability and precision in engine control.
38. The engine controller of claim 37 , wherein the one or more processors rely on an activation function.
The invention relates to an engine controller system designed to optimize engine performance by dynamically adjusting control parameters based on real-time data. The system addresses the challenge of improving engine efficiency, reducing emissions, and enhancing reliability by implementing adaptive control strategies that respond to varying operating conditions. The engine controller includes one or more processors configured to process sensor data from the engine, such as temperature, pressure, and fuel consumption metrics. These processors execute control algorithms to determine optimal engine settings, such as fuel injection timing, air-fuel ratio, and ignition timing. The system also incorporates a feedback mechanism to continuously monitor engine performance and adjust control parameters accordingly. A key feature of the invention is the use of an activation function to refine the control algorithms. The activation function processes input data, such as sensor readings or pre-defined thresholds, to generate output signals that influence engine control decisions. This function may be implemented as a mathematical model, a lookup table, or a machine learning-based approach, depending on the specific application. By integrating the activation function, the engine controller can achieve more precise and responsive adjustments, leading to improved engine operation under diverse conditions. The system may also include additional components, such as memory storage for storing historical data, communication interfaces for transmitting control signals to engine actuators, and diagnostic tools for identifying potential issues. The overall design ensures robust and adaptive engine control, enhancing performance while minimizing environmental impact.
39. The engine controller of claim 38 , wherein the activation function comprises one of the following: a hyperbolic tangent activation function; or Rectified Linear function.
This invention relates to an engine controller that uses machine learning to optimize engine performance. The controller includes a neural network with an activation function that processes input data, such as engine operating conditions, to generate control outputs. The activation function is designed to enhance the neural network's ability to model complex relationships between inputs and outputs, improving engine efficiency and emissions control. The activation function can be either a hyperbolic tangent function or a Rectified Linear Unit (ReLU) function. The hyperbolic tangent function outputs values between -1 and 1, providing smooth gradients for training, while the ReLU function outputs zero for negative inputs and the input itself for positive values, which helps mitigate vanishing gradient issues. The neural network processes sensor data, such as temperature, pressure, and speed, to adjust engine parameters like fuel injection timing and air-fuel ratio. The controller may also include a training module that updates the neural network weights based on real-time performance data to adapt to changing conditions. This approach improves engine control accuracy and adaptability compared to traditional rule-based systems.
40. The engine controller of claim 37 , wherein the one or more processors are arranged to learn on a training set of data using a learning technique.
This invention relates to an engine controller system designed to optimize engine performance and efficiency. The system addresses the challenge of adapting engine control parameters in real-time to varying operating conditions, which is critical for improving fuel economy, reducing emissions, and enhancing overall engine reliability. The engine controller includes one or more processors configured to execute control algorithms that dynamically adjust engine parameters such as fuel injection timing, air-fuel ratio, and ignition timing based on sensor inputs and predefined performance targets. A key feature of the controller is its ability to learn from a training set of data using a learning technique. This learning process allows the controller to refine its control strategies over time, improving accuracy and adaptability. The training data may include historical engine performance metrics, sensor readings, and environmental conditions, enabling the system to identify patterns and optimize control decisions. The learning technique could involve machine learning algorithms, statistical modeling, or other adaptive methods that enhance the controller's ability to respond to new or unforeseen operating scenarios. By incorporating learning capabilities, the engine controller can continuously update its control logic, ensuring optimal performance under diverse conditions. This adaptive approach reduces the need for manual calibration and improves long-term reliability. The system is particularly useful in modern engines where real-time adjustments are essential for meeting stringent emissions regulations and fuel efficiency standards.
41. The engine controller of claim 40 , wherein the learning technique includes one of the following: Limited-memory BFGS; or Stochastic Gradient Descent (SGD).
This invention relates to an engine controller that uses machine learning techniques to optimize engine performance. The controller is designed to address the challenge of adapting to varying operating conditions, such as changes in fuel quality, environmental factors, or wear and tear, which can degrade engine efficiency and emissions over time. The controller includes a learning module that continuously updates its control parameters based on real-time sensor data to maintain optimal performance. The learning module employs advanced optimization algorithms to adjust engine control parameters. Specifically, it uses either the Limited-memory BFGS (Broyden–Fletcher–Goldfarb–Shanno) method or Stochastic Gradient Descent (SGD) to refine the control strategy. The BFGS method is a quasi-Newton optimization technique that approximates the Hessian matrix to efficiently navigate the parameter space, while SGD is a gradient-based approach that iteratively updates parameters using small batches of data. Both methods enable the controller to adapt quickly to dynamic conditions without requiring extensive computational resources. The controller integrates with existing engine sensors to collect data on parameters such as fuel injection timing, air-fuel ratio, and exhaust emissions. The learning module processes this data to identify deviations from optimal performance and adjusts control signals accordingly. This adaptive approach ensures that the engine operates efficiently under varying conditions, reducing fuel consumption and emissions while maintaining reliability. The use of these optimization techniques allows the controller to balance computational efficiency with accuracy, making it suitable for real-time applications in automotive and industrial engines.
42. The engine controller of claim 27 , further comprising a misfire counter arranged to count a plurality of misfires as determined by the machine learning module and to generate a notification when the plurality of misfires exceeds a threshold value.
An engine controller monitors and controls internal combustion engine performance using a machine learning module that analyzes sensor data to detect engine misfires. The machine learning module processes input signals from engine sensors, such as crankshaft position, combustion pressure, or exhaust gas composition, to identify misfire events. The controller includes a misfire counter that tracks the frequency of these detected misfires. When the misfire count exceeds a predefined threshold, the controller generates a notification to alert the operator or maintenance system of potential engine issues. This notification can trigger diagnostic procedures, maintenance scheduling, or corrective actions to prevent further damage or performance degradation. The system improves engine reliability by proactively identifying and addressing misfire conditions before they escalate. The machine learning module may use historical data and real-time sensor inputs to refine misfire detection accuracy over time. The threshold value for misfire count can be dynamically adjusted based on engine operating conditions or predefined maintenance schedules. This approach enhances engine longevity and reduces unplanned downtime by integrating predictive analytics with traditional engine monitoring.
43. The engine controller of claim 27 , wherein the firing level modulation mode is a skip fire mode, wherein for a given reduced effective displacement that is less than full displacement of the internal combustion engine while operating in the skip fire mode, at least one cylinder is fired, skipped and selectively either fired or skipped in successive working cycles.
This invention relates to engine control systems for internal combustion engines, specifically addressing fuel efficiency and emissions reduction by modulating cylinder firing levels. The problem solved is optimizing engine performance under varying load conditions by selectively activating or deactivating cylinders in a controlled manner, rather than operating all cylinders continuously. The engine controller implements a skip fire mode, where the engine operates with a reduced effective displacement compared to full displacement. In this mode, at least one cylinder is fired in some working cycles and skipped in others, with the firing or skipping of the cylinder being selectively adjusted in successive cycles. This selective modulation allows the engine to adapt to changing power demands while minimizing fuel consumption and emissions. The system dynamically adjusts the firing sequence to maintain optimal efficiency without compromising performance. The skip fire mode ensures that the engine operates with fewer active cylinders when full power is not required, reducing unnecessary fuel consumption and mechanical wear. The selective firing and skipping of cylinders in successive cycles provides fine-grained control over engine output, enabling smoother transitions between different operating states. This approach improves overall fuel economy and reduces emissions compared to traditional fixed-displacement or full-cylinder deactivation strategies. The controller ensures seamless integration with existing engine management systems, allowing for real-time adjustments based on load conditions and driver input.
44. The engine controller of claim 27 , wherein the firing level modulation mode is a rolling skip fire mode where cylinders of the internal combustion engine are selectively fired or skipped in a predefined rolling pattern.
This invention relates to engine control systems for internal combustion engines, specifically addressing the challenge of optimizing fuel efficiency and reducing emissions by dynamically adjusting cylinder firing patterns. The system includes an engine controller configured to operate in a rolling skip fire mode, where individual cylinders are selectively fired or skipped according to a predefined rolling pattern. This pattern ensures that the engine maintains smooth operation while minimizing fuel consumption and emissions. The rolling skip fire mode involves a sequence where cylinders are activated and deactivated in a coordinated manner, allowing the engine to adapt to varying load conditions without abrupt power fluctuations. The controller monitors engine parameters such as torque demand, speed, and load to determine the optimal firing sequence, ensuring efficient performance across different operating conditions. By selectively deactivating cylinders in a rolling pattern, the system reduces unnecessary fuel consumption while maintaining drivability and responsiveness. This approach is particularly useful in applications where fuel economy and emissions reduction are critical, such as hybrid vehicles or engines operating under variable load conditions. The invention improves upon traditional skip fire methods by providing a more refined and adaptive control strategy, enhancing overall engine efficiency and performance.
45. The engine controller of claim 27 , wherein the firing level modulation mode is a dynamic multi-charge level mode where all cylinders of the internal combustion engine are fired, but individual working cycles are operated at different output levels.
This invention relates to engine control systems for internal combustion engines, specifically addressing the challenge of optimizing fuel efficiency and performance by dynamically adjusting the output levels of individual working cycles while maintaining continuous operation of all engine cylinders. The system includes an engine controller configured to operate in a dynamic multi-charge level mode, where each cylinder fires during every cycle, but the output level of each firing event is independently modulated. This approach allows the engine to produce varying power outputs without deactivating cylinders, thereby improving responsiveness and efficiency under partial-load conditions. The controller dynamically adjusts the charge level (e.g., fuel injection quantity or ignition timing) for each cylinder based on real-time operating conditions, such as load demand, speed, or emissions requirements. By distributing the workload across all cylinders at different output levels, the system reduces inefficiencies associated with traditional cylinder deactivation or fixed-output strategies. The invention is particularly useful in applications requiring smooth power delivery and rapid transient response, such as hybrid vehicles or engines operating in variable-load environments. The dynamic modulation of firing levels enables finer control over torque output and fuel consumption, enhancing overall system efficiency.
46. The engine controller of claim 45 , wherein the individual working cycles are operated at different output levels by using different air charge and/or fueling levels.
This invention relates to engine control systems designed to optimize engine performance by varying output levels across individual working cycles. The problem addressed is the need for precise control of engine output to improve efficiency, reduce emissions, and enhance responsiveness. Traditional engine control systems often operate at fixed or limited output levels, which can lead to inefficiencies and suboptimal performance under varying load conditions. The engine controller dynamically adjusts output levels for each working cycle by independently controlling air charge and fueling levels. This allows the engine to operate at different power outputs for each cycle, enabling finer control over torque delivery and emissions. The system may also incorporate feedback mechanisms to monitor and adjust these parameters in real-time, ensuring optimal performance under varying operating conditions. By decoupling air charge and fueling, the controller can achieve more precise combustion control, reducing fuel consumption and emissions while maintaining or improving power output. The invention is particularly useful in applications requiring variable torque delivery, such as hybrid or electric vehicles with internal combustion engine assistance, or engines operating under fluctuating load demands. The ability to independently adjust air charge and fueling for each cycle provides greater flexibility in engine management, allowing for smoother transitions between different operating modes and improved overall efficiency.
47. A method for controlling an internal combustion engine, the method comprising: operating cylinders of the internal combustion engine in a skip fire mode such that first firing opportunities of the cylinders are command to be fired while second firing opportunities of the cylinders are commanded to be not fired and intentionally skipped; using artificial intelligence to differentiate between (a) actual misfires of the plurality of cylinders that are commanded to be fired and (b) the intentional skips of the cylinders commanded to be not fired; and identifying the actual misfires differentiated from the intentional skips.
This invention relates to internal combustion engine control, specifically addressing the challenge of distinguishing between intentional cylinder skips and unintended misfires in skip-fire operation. Skip-fire mode selectively activates and deactivates cylinders to optimize fuel efficiency and reduce emissions, but this creates difficulty in detecting genuine misfires. The method operates an engine in skip-fire mode, where some firing opportunities are intentionally skipped while others are commanded to fire. Artificial intelligence is employed to analyze engine data and differentiate between actual misfires (unintended failures to fire) and intentional skips. The system identifies and isolates genuine misfires from the planned skips, enabling accurate diagnostics and engine performance optimization. This approach improves engine reliability by ensuring misfires are detected and addressed while maintaining the benefits of skip-fire operation. The AI-based differentiation enhances diagnostic accuracy, reducing false positives and ensuring proper engine function.
48. The method of claim 47 , wherein using artificial intelligence further comprises: measuring angular crankshaft acceleration of a crankshaft of the vehicle; receiving at a neural network a plurality of inputs indicative of operation of the vehicle while operating in the skip fire mode; using the neural network to predict angular acceleration of the crankshaft in response to the plurality of inputs; and determining when the misfires occur by comparing the measured angular acceleration with the predicted angular acceleration of the crankshaft.
This invention relates to engine control systems for vehicles operating in skip fire mode, where cylinders are selectively activated or deactivated to improve fuel efficiency. The problem addressed is detecting misfires in skip fire operation, which is challenging due to irregular combustion patterns. The solution involves using artificial intelligence, specifically a neural network, to predict crankshaft angular acceleration based on vehicle operating conditions. The system measures actual crankshaft acceleration and compares it with the neural network's predictions to identify misfires. The neural network processes multiple inputs related to vehicle operation, such as engine speed, load, and cylinder activation patterns, to generate accurate predictions. By analyzing discrepancies between measured and predicted acceleration, the system can reliably detect misfires even in skip fire mode, where traditional detection methods may fail. This approach improves engine performance and reduces emissions by ensuring proper combustion in active cylinders. The neural network is trained to account for the dynamic nature of skip fire operation, adapting to varying engine conditions. The method enhances diagnostic accuracy and enables real-time misfire detection, supporting advanced engine control strategies.
49. The method of claim 48 , wherein the one or more inputs are selected from the group including: spark timing; fuel mass; fire skip status; fire enable flag; cylinder skip number; order skip number; mass air per cylinder; cam phase timing; charge air temperature; engine speed; manifold absolute pressure; transmission gear; Deceleration Cylinder Cut-Off (DCCO) exit; vehicle speed; torque request; pedal position; fuel pressure; and turbocharger waste gate position.
This invention relates to engine control systems, specifically methods for optimizing engine performance and efficiency by dynamically adjusting various engine parameters based on real-time inputs. The problem addressed is the need for precise and adaptive control of internal combustion engines to improve fuel efficiency, reduce emissions, and enhance drivability under varying operating conditions. The method involves processing one or more inputs to determine optimal engine control settings. These inputs include spark timing, fuel mass, fire skip status, fire enable flag, cylinder skip number, order skip number, mass air per cylinder, cam phase timing, charge air temperature, engine speed, manifold absolute pressure, transmission gear, Deceleration Cylinder Cut-Off (DCCO) exit, vehicle speed, torque request, pedal position, fuel pressure, and turbocharger waste gate position. By analyzing these parameters, the system can adjust engine operations dynamically, such as modifying ignition timing, fuel injection, or cylinder deactivation, to optimize performance. The method ensures that the engine operates efficiently while meeting emissions standards and driver demands. This approach is particularly useful in modern engines where multiple variables must be balanced to achieve optimal results.
50. The method of claim 47 , wherein identifying the actual misfires further comprises generating a misfire flag when a misfire detection metric exceeds a predefined threshold, the misfire detection metric derived from a comparison of a measured angular acceleration with a predicted angular acceleration of a crankshaft of the vehicle.
This invention relates to engine misfire detection in vehicles, specifically a method for identifying actual misfires in an internal combustion engine. The problem addressed is the need for accurate and reliable misfire detection to prevent engine damage, reduce emissions, and improve performance. The method involves comparing a measured angular acceleration of the crankshaft with a predicted angular acceleration to detect misfires. When the difference between these values exceeds a predefined threshold, a misfire flag is generated, indicating a misfire event. The predicted angular acceleration is derived from engine operating parameters, such as speed, load, and fuel injection timing, while the measured angular acceleration is obtained from crankshaft position sensors. The method further includes filtering the measured acceleration signal to reduce noise and improve detection accuracy. By comparing the filtered measured acceleration with the predicted acceleration, the system can distinguish between actual misfires and transient disturbances. This approach enhances misfire detection reliability, enabling timely corrective actions to maintain engine efficiency and compliance with emissions standards. The invention is particularly useful in modern engines where precise control and monitoring are critical for performance and environmental compliance.
51. The method of claim 47 , further comprising: counting the identified actual misfires; and generating a notification when the counted misfires exceeds a threshold value.
This invention relates to engine misfire detection and monitoring in internal combustion engines. The problem addressed is the need to accurately detect and quantify misfires to prevent engine damage and ensure optimal performance. The system identifies misfires by analyzing engine speed fluctuations, which are indicative of incomplete combustion events. Specifically, the method involves measuring crankshaft speed variations, comparing these variations to a predefined threshold, and identifying misfires when the variations exceed the threshold. The invention further includes counting the detected misfires and generating an alert when the count surpasses a predetermined threshold value. This notification can trigger maintenance actions or adjustments to engine operation to mitigate potential damage. The system may also adjust the misfire detection threshold dynamically based on engine operating conditions, such as load or speed, to improve accuracy. The method ensures reliable misfire detection across different driving conditions, enhancing engine longevity and efficiency.
Unknown
October 27, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.